Simplifying deflation for non-convex optimization with applications in Bayesian inference and topology optimization
This work addresses a bottleneck in non-convex optimization for fields like Bayesian inference and topology optimization, though it appears incremental as it builds on existing deflation concepts.
The paper tackles the problem of exploring multiple local optima in non-convex optimization by proposing a simple, general deflation constraint that integrates with existing solvers, with applications demonstrated in Bayesian inference and topology optimization.
Non-convex optimization problems have multiple local optimal solutions. Non-convex optimization problems are commonly found in numerous applications. One of the methods recently proposed to efficiently explore multiple local optimal solutions without random re-initialization relies on the concept of deflation. In this paper, different ways to use deflation in non-convex optimization and nonlinear system solving are discussed. A simple, general and novel deflation constraint is proposed to enable the use of deflation together with existing nonlinear programming solvers or nonlinear system solvers. The connection between the proposed deflation constraint and a minimum distance constraint is presented. Additionally, a number of variations of deflation constraints and their limitations are discussed. Finally, a number of applications of the proposed methodology in the fields of approximate Bayesian inference and topology optimization are presented.